Anyone ever tried binaural beats to test stereo separation?

The way to do it is create a steady sine wave (say 1 KHz) in one channel, and a steady sine wave 4 or 5Hz higher or lower in the other.

If you wear headphones, there will be a throbbing or beating effect, and it seems to me that if stereo separation is not good at some given frequency, the beat effect will be reduced.

What does everyone (anyone?) think of this for a future testing method... ? Or has it already been tried?

I haven't specifically tried this lately, but it has been used in the past in various situations by other LAME developers I believe. Frank Klemm has also experimented with artificially generated sound samples before with a similar nature.



As with all contrived situations, it's easy to create a sample which will purposely trip up an encoder. One can create samples that do this for even the best encoders/formats such as ogg, mpc, aac (psytel or fhg), and LAME. Artificially generated samples then, don't have much value in determining general quality. Instead they can only be used to make absolute statements such as "LAME is not perfect".... but we already know the answer to these questions most of the time

I think it's much more valuable to find multiple samples in real music which show artifacting for whatever reason (in this case stereo related) because the implications for general behavior are much greater in that sense.

So in short, it may be useful for experimentation, but its probably not so useful for overall quality assessment or tuning. There are some exceptions to this case though, but that's usually when a particular psymodel in question is still in an early stage and still going through very active development.

Thanks Dibrom... one thing though, in some new age music, binaural beats are actually used as part of the "atmosphere" of the music, as they tend to entrain brainwaves and create a feeling of peace or hypnosis. One example (artist) I can think of is Kelly Howell, and also the Monroe Institute's "Hemi-Sync" stuff. So it's conceivable that some music may actually be using artificial methods... other electronica type stuff comes to mind. Some of that may get encoded into MP3 format for whatever reasons...

I do understand what you're saying though about "purely" artificial samples designed to trip up an encoder. Sure, perfection would be impossible in anything but lossless encoding.

Cheers,

fewtch (Tim)

If anyone wants to try this, I uploaded a file at:

http://ff123.net/export/throb.pac

The left channel is 1 kHz, and the right channel is 1.005 kHz (both 9dB down from full scale).

When I encoded with the buggy FastEnc, it sounded more "throbby" than the original. Using --alt-preset standard, it sounded the same to me.

ff123